Core i7 7700K vs. Ryzen 7 1800X With Ubuntu 17.04 + Linux 4.12

[bug77]

it does not matter. as long as 6900k exists and has same speed as 1800x, 1800x competes with 6900k. if you are trying to argue that 6900k has no place in this world, talk no intel, not no me

Well, if you're including stuff just because it exists and runs at the same clock speed, don't forget about P4 560 or Pentium D 960. Guaranteed to make Ryzen look like a steal.

[pal666]

i7-6800K is in the same price range

it does not matter. as long as 6900k exists and has same speed as 1800x, 1800x competes with 6900k. if you are trying to argue that 6900k has no place in this world, talk no intel, not no me

[caligula]

i7-6800K is in the same price range and is very competitive. In fact, it beats 1800X in most common use cases. What matters is real world performance and price, not theoretical specifications. The same argument was used for Bulldozer; it had more cores so it had to be "better".

So what are the most common use cases for people buying 6+ core CPUs? Definitely not games. The best gaming rigs use 4 or fewer cores.

[efikkan]

i messed it up, competition is i7-6900k, which is of more or less same speed and two times more expensive
6800 is competition for 6core ryzens

i7-6800K is in the same price range and is very competitive. In fact, it beats 1800X in most common use cases. What matters is real world performance and price, not theoretical specifications. The same argument was used for Bulldozer; it had more cores so it had to be "better".

[fuzz]

"CPU with higher clock speed goes faster." <-- not true, depends on the task. All Intel server CPUs with 12+ cores have much lower clock speed and they 'go faster' in those tasks they were designed for. Game FPS isn't the only metric. Throughput is much better metric for some useful tasks. Besides, with high throughput, you can often even finish tasks faster for example in web servers. It just depends on how you define tasks.

Take your foot out of your ass and take off your flame colored glasses. We agree. Maybe you should learn English.

[pal666]

The competition is i7-6800K, which is slightly cheaper.

i messed it up, competition is i7-6900k, which is of more or less same speed and two times more expensive
6800 is competition for 6core ryzens

[efikkan]

pointless comparison. 1800x's competition is $1100 6950
if you feel urge to compare 7700, compare it against 6950, not against two times cheaper ryzen

The competition is i7-6800K, which is slightly cheaper. i7-6800K beats it in most workloads, including gaming, CAD, photo editing and more.

[bug77]

Not to mention that even mobile phones have moved to 8 core CPU's. More cores is the only practical way to achieve significant performance gains.

[torsionbar28]

You're still sceptical even though ESP32, $9 RPi clones, and anything upwards has multiple cores and the number is increasing with each year? A quad-core ARM costs ~ $2. Everyone can already buy them.

Not to mention that even mobile phones have moved to 8 core CPU's. More cores is the only practical way to achieve significant performance gains.

[caligula]

It puzzles me to see people thinking multithreading is the new kid on the block that will save puppies from dying one we figure out how to properly put it to good use. The reality is we've had (super)computers that would run hundreds of threads for decades. We've written code for them.

That's BS or applies to a very specific definition of 'we', the HPC community. Normal people had one core/active thread/CPU per computer till ~2005. Some professionals had dual socket workstations. Some IBM Power machines had SMT before Intel, that is two hardware threads. Even HPC machines might have had one core/CPU/active hardware thread per machine. The supercomputers were normal boxen with fast interconnect LAN. There are some exceptions, like the SUN 'barrel' processor. Tilera and other research platforms were not available for consumers.

Another thing, the server software behaves differently. The old servers from 1990s and 2000s used to run on a cluster of machines. One machine as reverse proxy, few as DB servers, a dozen as frontend servers and so on. Now the slow clusters (with 1/10G eth interconnect) are being replaced by heavy 22+ core CPUs on the cloud. So, even now the multi-socket and multicore are way faster than the old stuff and it truly is an upgrade. PCIe and QPI/UPI are way faster than gigabit ethernet.

But the simple fact is not all problems are infinitely parallelizeable and even when you find tasks that are, you're hit by other problems that impact multithreading (e.g. memory coherence - look what happens the your data resides in another CCX's cache).

I've used dual socket since Pentium 2. It worked just fine for tasks like compiling stuff on Gentoo (well, I used Slackware back then, but anyways). The first gen of Intel dual-core had a similar legacy design with two totally separate CPUs on the same die. When they integrated both as cores in Core2Duo, the main advantage was the shared cache and a bit faster inter-communication. However, most tasks like make -j2 don't get any magical boost from tighter core integration. Compiling scales so well that distcc speeds it up even on a 100 Mbit network. The core/uncore design with shared caches helps with the coherence a bit, but cache coherence won't become a problem anytime soon. At current rate, it will take Intel 10 years to deliver the core counts from their heavyweight server segment to consumers. The current designs already show that they work just fine.

My recommendation is stop trying to predict the future. Buy what works for you now and let people in the know handle the advancements. When/if advancements come, upgrade and enjoy. When/if advancements don't come, enjoy your current rig

It's pretty clear that there's only one way up, more cores and wider SIMD. If you look back, the biggest advancements in perf scaling come from SSE/AVX/cores. Faster clocks and IPC come far behind. BTW, guess how they got IPC up, that's right by adding more ALU/FP pipelines. IBM's fastest stuff runs at 5 GHz, Intel at 4.5 GHz. Next Intel i7 8700k will have 4 cores with 2% more speed? We'll see. People could already overclock their system 2% just like that. With a jump to 6..8 cores, they can achieve more than any IPC/freq boost would ever do.

This whole multicore craze reminds me of the early 2000s when people where convinced that since they suddenly started thinking about EVs, by 2005 everyone will be driving one.

You're still sceptical even though ESP32, $9 RPi clones, and anything upwards has multiple cores and the number is increasing with each year? A quad-core ARM costs ~ $2. Everyone can already buy them.